Skip to main content
vote digital omnibus ai actreport ai act décembre 2027imco libe vote 2026

IMCO/LIBE Vote of March 18: European Parliament Backs Delay of AI Act to December 2027

18 March 20265 min3
IMCO/LIBE Vote of March 18: European Parliament Backs Delay of AI Act to December 2027

Key takeaways

  • Vote adopted March 18, 2026: IMCO and LIBE committees adopted their joint position by 101 votes in favour, 9 against and 8 abstentions
  • Fixed dates, not conditional: unlike the Commission's original proposal, Parliament introduces firm deadlines — December 2, 2027 for Annex III, August 2, 2028 for Annex I
  • Watermarking (Art. 50§2): November 2, 2026 — Parliament shortens the grace period proposed by the Commission (which was February 2027)
  • New ban: "nudifier" applications generating sexually explicit images without consent are added to the list of prohibited practices
  • Small mid-caps confirmed: companies with up to 750 employees and €150M turnover will benefit from the same reliefs as SMEs
  • Next step: plenary vote on March 26 — then trilogue negotiations with Council. August 2026 remains the legal deadline until final adoption

What we anticipated in our Digital Omnibus analysis has just cleared a decisive milestone. On March 18, 2026, the IMCO (Internal Market) and LIBE (Civil Liberties) parliamentary committees adopted their joint position on the Digital Omnibus AI Act by 101 votes in favour, 9 against and 8 abstentions. A strong political signal in favour of the delay — but the text is not yet law.

What Parliament Actually Voted On

Wednesday's vote concerns the European Parliament's position ahead of trilogue negotiations with the Council. Several points deserve close attention, as they diverge from the Commission's original proposal.

Fixed dates instead of the conditional mechanism

The Commission had proposed a conditional mechanism: obligations would apply 6 or 12 months after it confirmed the availability of harmonised standards. Parliament prefers legal predictability and introduces firm deadlines:

  • December 2, 2027 for high-risk systems under Annex III (biometrics, critical infrastructure, education, employment, essential services, law enforcement, justice, border management)
  • August 2, 2028 for systems embedded in products regulated by sectoral legislation (Annex I — medical devices, machinery, radio equipment…)
"We want predictable, stop-the-clock, simplified rules that remove overlaps with sectoral legislation and reduce fragmentation between Member States. Companies need clarity on whether they are high risk or not." — Arba Kokalari, IMCO co-rapporteur (EPP, Sweden)

Watermarking: a shorter grace period than expected

On Article 50§2 (machine-readable marking of AI-generated content), Parliament is more restrictive than the Commission. Where Brussels proposed a deadline of February 2, 2027 for systems already on the market, MEPs propose November 2, 2026 — three months earlier. Other Article 50 transparency obligations continue to apply from August 2026.

New ban: "nudifier" applications

Parliament introduces an addition not included in the Commission's proposal: an explicit ban on "nudifier" systems — applications that use AI to generate or manipulate sexually explicit images depicting an identifiable real person without their consent. This ban is added to the list of prohibited practices under Article 5. Exception: systems with effective safety measures preventing such use.

What Hasn't Changed Since Our Previous Article

The main pillars of the Digital Omnibus remain intact:

  • Extension of reliefs to small mid-caps (under 750 employees, under €150M turnover) confirmed
  • Strengthening of the AI Office as the central supervisory authority
  • Possibility of processing sensitive data to detect and correct biases, under strict safeguards
  • Reduction of overlaps between the AI Act and existing sectoral legislation
  • Prohibited practices, GPAI obligations and basic transparency (Art. 5, 51–56, 50 excluding §2) remain unaffected

Next Steps Before Final Adoption

This committee vote is a milestone — not the finish line. The timeline ahead:

  1. March 26, 2026: plenary vote to validate the committees' negotiating mandate
  2. Spring 2026: trilogue negotiations between Parliament, Council and Commission
  3. Mid-2026: expected final adoption — under pressure from the August 2026 deadline

Until that final adoption, August 2, 2026 remains the legally binding date. If trilogues run late and the text is not adopted before that date, the original AI Act obligations apply automatically.

The strategy remains unchanged: prepare as if August 2026 is real, plan as if December 2027 is the likely enforcement date. The free AiActo diagnostic lets you classify your AI systems and identify your obligations right now — whatever the final date turns out to be.

Frequently Asked Questions

Does the March 18 vote mean the delay is guaranteed?

No. The March 18 vote is the position of the IMCO and LIBE parliamentary committees. It still needs to be validated in the plenary on March 26, then negotiated with the Council during trilogues. The final text can still be amended. Definitive adoption is not expected before mid-2026.

What's the difference between the Commission's proposal and Parliament's position?

The Commission proposed a conditional mechanism where obligations would start 6 or 12 months after a decision confirming the availability of standards, with backstops at December 2027 and August 2028. Parliament simplifies this by setting those backstops directly as fixed application dates — December 2, 2027 for Annex III and August 2, 2028 for Annex I — providing greater legal certainty for businesses.

Why did Parliament shorten the watermarking grace period?

Parliament proposes November 2, 2026 (instead of the Commission's February 2, 2027) for machine-readable marking of AI-generated content (Article 50§2). This reflects a political will to maintain pressure on synthetic content transparency, which several political groups consider a priority.

What exactly is the new ban on "nudifier apps"?

Parliament proposes adding to the list of prohibited practices (Article 5) AI systems that generate or manipulate sexually explicit or intimate images depicting an identifiable real person without their consent. The ban does not apply to systems with effective safety measures preventing such use.

The March 18 vote marks a major milestone: for the first time, the European Parliament has officially backed delaying AI Act high-risk obligations with firm dates. But the legislative process continues — and August 2026 remains the legal deadline until proven otherwise. Check the AI Act timeline on AiActo to follow regulatory calendar updates in real time.

Share this article